Professional Documents
Culture Documents
BRIAN HOPKINS
{Botany Department, Unmerrity College, London)
J. G. SKELLAM
(The Nature Conservancy, London)
WITH AN APPENDIX BY
214
3. CURRENT METHODS
.3.1. All the current methods involve the counting of the number of
individuals present in quadrats which are either thrown at random or arranged
in a regular manner.
3.2. If the individuals are randomly distributed, then the proportion of
random quadrats containing o, 1, 2, 3, . . . individuals will be given by the
values of the terms of a Poisson series
e~m,
merm,
m%e-m\z\,
OT3^-"1^
!,...,
all sizes (Clements states that the area need not be i sq. m.) and shapes, so
that it would seem justifiable to define it as above.
2.2. Distribution is static arrangement in space; dispersion is a dynamic
concept, implying dissemination. Dispersed, being the past tense, may be
applied in some cases (e.g. in the terms overdispersed and underdispersed) to
the results of dispersion.
Three types of distribution have been distinguished, random, aggregate,
and regular. A random, or 'normal', distribution is one in which the individuals
of a population are distributed according to chance. There has been considerable confusion in the terminology of non-random distributions. This
has been mainly due to the use of the term overdispersed in two opposing
meanings. The statistical term of Svedberg (1922) refers to the distribution
of quadrats containing o, 1, 2, 3, . . . individuals; his overdispersed population
has more empty quadrats than would be expected by chance, and vice versa
for his underdispersed population. Also there is a physical meaning referring
to the actual distribution of the individuals on the area; an overdispersed
population has its individuals evenly distributed, and an underdispersed
population has its individuals in groups. Because of this confusion, the terms
overdispersed and underdispersed will not be used in this paper.
The term overdispersed used in its statistical sense (grouping of individuals)
will be replaced by the term aggregated as suggested by Goodall (1952), and
the term underdispersed used in its statistical sense (evenly distributed
individuals) will be replaced by the term regular.
215
216
0-6
0-5
0-4
0-3
20
30
40
50
n
FIG. 1. Abac for determining the probability (p) of exceeding
the calculated value of x for n observations by chance.
Nine different populations were used. The details of these are given in
Table I.
In the aggregated populations the individuals were placed in hexagons of
side 25 mm. containing seven individuals, the central one being placed at
random. In the regular distributions the individuals were placed at random
except where the position of an individual fell within 25 mm. of one which
had already been placed. When this happened, a new random grid reference
was selected, and the process repeated until the required number of individuals
has been placed.
0-7
217
TABLE I
Type of distribution.
Random
Total density.
A i
Az
A3
,,
,,
200
Bi
Bz
B3
Aggregated
400
IOO
,,
200
400
Regular
Cz
,,
,,
IOO
200
400
Density per (
0-25
050
i-oo
025
050
i-oo
025
0-50
i-oo
6.I. Methods. The method used for natural populations was similar to
that used for the synthetic populations. A uniform area of a plant population
C i
c3
IOO
Relative variance.
Population A i
i
1-0528
0-9899
4
16
25
100
0-8125
0-8640
1 -0667
N rmal
deviate.
075
0-07
065
420
082
0-4-0-5
> 09
0-5-0-6
0-7-0-8
0-4-0-5
230
0-02-0-05
464
2232
036
0-7-0-8
104
0386
O-II
123
> 0-9
0-2-0-3
0-2-0-3
23
0-37
X*
98
195
0752
0036
P
0-05-0-06
0-96-0-97
0-2-0-3
130
32
0-3-0-5
0-3-0-5
Population A 2
1
4
16
25
100
1-1630
1-0505
0-9688
1-4507
2-2534
Population A 3
1
1-0576
1-0404
4
16
25
100
1-0417
0-9280
0-5600
Population B i
1
26968
5-7980
14
16
10-2708
6-7733
25
100
27467
i-54
081
028
0-14
O-2O
054
2397
3376
32-n
1581
214
0-4-0-5
07-0-8
0-8-0-9
0-8-0-9
0-6-0-7
<
<
<
<
o-ooi
o-ooi
o-ooi
o-ooi
O-I-O-2
21-8
6-8
422
103
0-821
0-317
246
IOI
82
OO
X*
08580
1-2608
0-02-0-03
070-071
0-3-0-5
O-I-O-2
0-05-0-10
1-5089
5-5972
0-41-0-42
0-8990
5-8609
075-076
Coefficient of
aggregation.
d.f.
0-3-0-5
0-5-07
08655
80
P
0181
1
<;
2
0-3-0-5
0-2-0-3
1-1842
86
0-133
3
7
0-80-9
0-5-0-7
1-0401
839788
1
2
< o-ooi
< o-ooi
161499
909395
94
O-394
<< o-ooi
<^o-ooi
< o-ooi
< o-ooi
0-02-0-05
5
8
!
0-3-0-5
18-149
19-846
1
cF
g
0-3-0-5
0-3-0-5
25
139
17
1,076
574
N
M
Fit to Poisson.
1
VjM
TABLE II
S9
< o-ooi
25
56534
2609
3177
2850
1275
29467
238
Population B 3
1
2-6416
4-4849
4
16
52552
25
3 5253
IOO
3-8867
2319
2452
1474
6-92
354
IOO
Population C i
1
0-8320
O7474
4
16
08125
0-9067
25
IOO
0-4800
<
<
<
<
1,136
o-ooi
o-ooi . 546
222
o-ooi
85
o-ooi
0-I-O-2
178
065
12
O-OOI-O-OI
0-0I-0-O2
0-05-0-10
332
351
< b-ooi
2-63
i-59
O-OOI-O-OI
25
0-9600
O-II
IOO
O-4934
0-62
0-5-0-6
5 92
398
2-35
o-8o
< o-ooi
< o-ooi
4
16
25
IOO
c3
05815
0-4344
03229
07093
01333
106
o-i-o-a
>
0-02-0-05
53
Population C 3
1
O7518
4
0-6263
16
05417
Population
8-8
< o-ooi
-< o-ooi
< o-ooi
< o-ooi
0-02-0-05
o-ooi
o-ooi
o-ooi
o-ooi
0-5-0-6
07-0-8
0-5-0-6
0-26
0-64
19-010
< o-ooi
^ o-ooi
< o-ooi
< o-ooi
<
<
<
<
2-37
19435
09
0-02-0-05
0-4-0-5
0-3-0-4
1.054
444
17682
15-763
126
74
-2463
-1-870
300
62
13
144
i-5
232
43
77
io-6
0-4
2-900
169-8342
652-8012
< o-ooi
< o-ooi
7-1303
73
o-ooi
< o-ooi
< o-ooi
49640
89
o-ooi
O-OOI-O-OI
0-2-0-3
0-9342
72
0341
0-7835
94
0-026
05563
89
68861
o-o6-o-o7
2-9347
14-6422
10-1179
< o-ooi
O-OOI-O-OI
23-1811
332879
3
7
< o-ooi
< o-ooi
0-3-0-5
0-3-0-5
-3736
228-6226
235-5285
O-OI-O-02
0-2-0-3
195
136
i-4
Population Ba
i
3-8472
55152
4
16
9-2292
< o-ooi
O-OOI-O-OI
0-02-0-05
0-5-07
0-3-0-5
6690
-4762
<< o-ooi
< o-ooi
< o-ooi
0-2-0-3
0-05-0-10
<
0001
dominant species; Narthecium ossifragum, Eriophorum, angustifolium, Rhynchospora alba, Schoenus mgricans, Molinia caerulea, and several species of
Sphagnum were very frequent. There were occasional pools and hummocks
on the site, which was similar to, but wetter than, those described by Pearsall
and Lind (1941). Fruiting heads of Eriophorum were chosen as individuals.
P. II. MAYO BOG (Utricularia intermedia) (July 1951). The site was
identical with that of P. I; the two analyses were done simultaneously. Flowering heads of Utricularia were chosen as individuals. The Utricularia grows
in bog pools, but only four of the pools on the site contained it, hence it
occurred in four clumps. This population was chosen as being an example of
apparent extreme aggregation.
P. III. RANNOCH BOG (Eriophorum angustifolium) (August 1951). The
site was quite close to Rannoch Station, Perthshire, on an area of bog at an
altitude of 950 ft., which had probably been burnt a few years before it was
sampled. There was no dominant species, and the most frequent species were
Calluna vulgaris, Erica tetraUx, Narthecium ossifragum, Trichophorum caespitosum, Eriophorum angustifolium, several species of Sphagnum, several Hepaticae,
22o
HopkinsA New Method for determining
was chosen and a 20 X 20 square grid set up, the length of the side of the grid
unit (L.G.U.) depending on the density of the individuals. The selected
area was the site within which all sampling was done. Every fifth individual
encountered, whilst traversing the rows to record the number of individuals
in each grid unit, was used as an individual from which the distance to its
nearest neighbour was measured.
The setting up of a grid is not necessary in order to determine the type of
distribution of a population. A series of lines are placed across the site and
every/th individual occurring within a short but fixed distance of one or both
sides of the line is used as an individual chosen at random.
Random points were chosen by throwing a cane at random. All distances
were measured at ground level in the herbaceous populations and at breast
height in the tree populations. No allowance was made for the size of the
individuals; the distance between them was measured to the nearest centimetre.
6.2. Populations studied. Twelve natural populations were analysed. Most
of these were from areas known to have been little altered by man, and thought
to have remained unchanged for hundreds of years. Hence the distribution of
the individuals was assumed to have attained equilibrium with the environment and to be entirely natural. They were selected to obtain a wide range
of type of distribution, habitat, and vegetation. However, as the new method
will potentially have its greatest use in wooded areas, more than half the
examples were chosen from tree populations.
A brief comment on each population will now be given.
P. I. MAYO BOG (Eriophorum angustifolium) (July 1951). The site was
4 miles south of Louisburgh, near Westport. It was part of a vast expanse of
blanket bog which had presumably been growing throughout the subatlantic
period and had apparently been little influenced by man. There was no
221
222
TABLE III
No.
Locality.
I
II
III
IV
V
VI
VII
Mayo Bog
Mayo Bog
Rannoch Bog
Hillock Wood
Ridley Wood
Rothiemurchu3 Wood
Cressbrookdale Wood
Watendlath Wood
Watendlath Grassland
South Bentley Wood
Cothill Fen
Kingley Vale Wood
VIII
IX
X
XI
XII
Species.
Area
Density
of
per
site
Total
L.G.U.
grid
(m.)
(sq. m.) density. unit.
400
400
400
10,000
40,000
3,600
900
6,400
510
IOI
91
382
193
347
SS5
106
100
213
240
100
35i
900
141
1-275
0-253
0-228
-955
0480
0868
1-388
0-127
-533
o-6oo
0-878
-353
The results do not require much comment. The Schoenus tussocks (P.
XI) were regularly distributed. The beech trees (Ridley Wood, P. V),
bracken fronds (P. IX), and oak saplings (P. X) were distributed at random.
The remaining eight populations were aggregated. They were, in order of
increasing aggregation, beech trees (Hillock Wood, P. IV), pine trees (P. VI),
birch trees (P. VIII), yew trees (P. XII), fruiting heads of Eriophorum angusti-
223
folium (Rannoch Bog, P. Ill, and Mayo Bog, P. I), ash shoots (P. VII), and
finally,floweringheads of Utricularia intermedia (P. II).
7. CONCLUSIONS
8. SUMMARY
As will be seen from Tables II and IV, results obtained by the new method,
described in this paper, agree closely with those obtained by the methods in
current use. The agreement is such that the probabilities of the significance
of the deviation of A from unity are of the same order as those for VjM and x2.
Thus the new method may be used as an alternative to the quadrat methods.
When analysing the type of distribution of trees using the current methods,
quadrats have to be marked out on the ground; this is tedious and time
consuming. The new method described in this paper is very quick; only
distances need to be measured. The line samples used for obtaining random
individuals take a little time but are much quicker than the use of quadrats,
especially where the number of individuals in each has to be counted. Hence
this new method would seem to be a useful substitute for the quadrat methods.
The interest of aggregated populations lies in the reason for and the type of
aggregation. The method described above only tells whether or not a population is distributed at random, and how different from random is its distribution. Work is proceeding relating P and I to the density of the individuals
and to the distribution patterns.
V/M
t
<
<
<
<
O-OOI
o-ooi
o-ooi
o-ooi
< o-ooi
X1
"t O-OOI
deviate
P
663
291
8183
10-089
< o-ooi
< O-OOI
128
61
10,364
2,516
< o-ooi
< o-ooi
"574
56-900
< O-OOI
< o-ooi
73
721
270
163
142
41
9742
9-202
1-265
9245
0-2-0-3
< o-ooi
< o-ooi
< o-ooi
< o-ooi
435
0-05-0-10
0-05-0-10
O-OS-O-IO
> 0-9
0-2-0-3
450
271
t o-ooi
< 0001
< O-OOI
d.f.
85-5021
38-8340
< o-ooi
101-8206
1447795.
I
2
< o-ooi
< o-ooi
33-3673
257258
I
2
< O-OOI
48-9776
16-6490
3
5
<
A
3-3632
97
< o-ooi
20
<t o-ooi
2-1990
19
0008
1-2610*
70
0-085
1-0318*
35
0434
0001
262678
< o-ooi
< o-ooi
< O-OOI
37
"7
60
73
10-3
4-8
6-4
< O-OOI
"< O-OOI
< o-ooi
714
531
Coefficient of
aggregation
Fit to PoUton
Relative variance
TABLE IV
1769
2-OO2
0-20-0-21
<
<
<
<
o-ooi
o-ooi
o-ooi
o-ooi
0-07-0-08
0-04-0-05
< o-oi
0-5-07
0-05-0-10
2-2872
89589
< O-OOI
O-OOI-O-OI
0-3-0-5
0-02-0-05
7-3231
37-0631
3-656
4-019
O-I-O-2
477
163
49
33
9
2,032
1,480
1,177
947
484
35-518
4O-37O
1,028
145
399
27-2
8-6
17-113
3-993
409
0-370
1-130
0-71-0-72
0-25-0-26
0-2-0-3
0-05-0-10
O-I-O-2
0-3401
1-4213
-0663
1-651
0-50-0-51
o-oo-o-io
0-3-0-5
< O-OI
0-05-0-10
0-3371
43373
I-IO6I
377
1-6505
4-58
< o-ooi
I*
25
2-0503
3-1917
3-64
3-36
O-OOI-O-OI
O-OOI-O-OI
O-OOI-O-OI
IOO
38780
3-30
1-4675
3-29
100
i-66o8
1-8113
3-8550
2-29
3-33
3-37
O-OOI-O-OI
0-030-05
0-03-0-05
O-I-O-3
Wood (Quercus)
0-67
0-5-0-6
i-59
0-1-0-2
O-2O
O-8-O-9
i-8o
0-05-0-10
i-oo
0-3-0-4
< o-ooi
< o-ooi
o-oi-o-oa
0-05-0-10
0-4-0-5
1-4471
3-15
O-OOI-O-OI
16
2-4081
5-09
< o-ooi
25
2-1711
3-21
O-OOI-O-OI
100
3-8487
3-49
0-02-0*05
152
29
5 9
42
i*o
418
143
59
33
13
< o-ooi
o-ooio-01
O-OOI-O-OI
0-02-0-05
i8-3
8-5
i-o
380
77
25-4
47
0-4
O-OOI-O-OI
-10795
-6-381
0-683
2-876
<t o-ooi
< o-ooi
< o-ooi
< o-ooi
< o-ooi
462-8216
271-1048
< o-ooi
6-1639
36333
O-OOI-O-OI
0-02-0-05
0-02-0-05
0-02-0*05
4 o-ooi
< O'OOI
< O-OI
< O-OI
O-I-O-2
O-49-O-5O
O-OOI-O-OI
< O-OOI
O-OOI-O-OI
O-OOI-O-OI
163-1565
43-1685
0-3833
10-2857
0-02-0-05
< O-OOI
1-3135*
69
0-055
< O-OOI
< O-OOI
5-4954*
in
<t o-ooi
O-OI-O-O3
0-I-O-3
i-69i5 f
31
0-044
0-8-0-9
0-8-0-9
1-1653
43
0-333
0-8-0-9
0-3-0-5
09386
48
0-348
< o-ooi
0-3163
70
< o-ooi
1-8751
38
o-oi 1
< O-OOI
0-5-07
O-OI-O-O2
226
APPENDIX
By J. G. SKELLAM
(The Nature Conservancy, London)
CLAPHAM, A. R., 1936: Overdispersion in Grassland Communities and the use of Statistical
Methods in Plant Ecology. J. Ecol., xxiv. 232.
CLEMENTS, F. E., 1905: Research Methods in Ecology, p. 321. Univ. Pub. Co., Lincoln,
U.S.A.
DAWKINS, C. J., 1939: Tussock Formation by Schoema rrigricans: the Action of Fire and Water
Erosion. J. Ecol., xxvii. 78.
FISHER, R. A., 1950: Statistical Methods for Research Workers, pp. 57-61. n t h edit. Oliver
and Boyd, Edinburgh and London.
GOODALL, D. W., 1952: Quantitative Aspects of Plant Distribution. Biol. Rev., xxvii. 194.
GREIG-SMITH, P., 1952: The Use of Random and Contiguous Quadrats in the Study of the
Structure of Plant Communities. Ann. Bot., xvi. 293.
Moss, C. E., 1913 : The Vegetation of the Peak District, pp. 65-74. Cambridge Univ. Press
PKARSALL, W. H., 1950: Mountains and Moorlands, pp. 129-31. Collins, London.
and LIND, E. M., 1941: A Note on a Connemara Bog Type. J. Ecol., xxix. 62.
PEARSON, K., 1914. Tables for Statisticians and Biometricians, Cambridge University Press.
1934- Tables of the Incomplete Beta-Function. Biometrika, London.
SKELLAM, J. G., 1952: Studies in Statistical Ecology, I. Spatial Pattern. Biometrika, m i x .
346.
SVEDBERG, T., 1922: Statistik vegetationsanalys. Svensk bot. Tidskr., xvi. 1.
TANSI^Y, A. G., 1949: The British Isles and their Vegetation, pp. 447-450. Cambridge
University Press.
WATT, A. S., 1926: Yew Communities of the So ith Downs. J. Ecol., xiv. 282.
1934: The Vegetation of the Chiltern Hills with Special Reference to the Beechwoods
and their Serai Relationships. Ibid., xxii. 237 and 445.
1947: Contributions to the Ecology of Bracken (Pteridium aqutlinum), XV. The Structure
of the Community. New Phytol., xlvi. 97.
227
Kr
is 1e~ **$. The probability that the nearest particle to the centre lies between r2 and rlt is
By letting rx-^>-r% = r and calling r%rx = dr, we obtain the probability
that the nearest particle to the centre lies at a distance between r and r-\-dr.
This is
*** zXrdr.
J2. We now make use of the following theorems, proofs of which are given
for example in Weatherburn (1947). 'A First Course in Mathematical
Statistics', chap. 8 (Cambridge University Press).
z
THEOREM I. If the variate z has the frequency function / (z) = e~ , then
the variate Z defined as the sum of n independent values each distributed as
z has the frequency function
Here
h(Z) = e-z^ir(n).
T(n) = (n1) (n2) ... 3. 2. 1.
(2)
x = ZJ^+Zt)
has the frequency function
j{x) = ^(i-xr*IB(n,n),
where o < x < 1 and B(n, n) = T(n) I\n)jl\2n).
(3)
3. If from n random points the distances to the nearest particle are r1(
r2, ..., rn, and if from n random particles the distances to the nearest neighbouring particles are pv p2, , pn, then both Zx = A ^ 2 and Z2 = A^p2 have
the distribution (2) by reasons of relation (1) and Theorem I.
It follows from Theorem II that
x
/ (z)dz = e~zdz.